The R-tree is a highly balanced tree, and is the most widely used spatial index structure. Based on the similarity of the historical data, this paper constructs R-tree, and proposes a collaborative filtering recommendation algorithm based on R-tree (R_CF). In addition, this paper sets about from the user's implicit feedback, builds the user interest behavior data model, and standardizes data. Simulation experiments show that compared with the traditional collaborative filtering recommendation algorithm (CF), the proposed R_CF algorithm can greatly improve recommended top-n query speed.